Drip Saves Big on Black Friday CRM Push with Amazon EC2 Spot Instances

Before a business can make a sale, it must make a connection—usually through word of mouth, advertising, or direct communication. Startup e-commerce customer relationship management (CRM) firm Drip helps businesses manage their connections by sending emails on behalf of its retail customers—and at no time are these connections more important than in the days leading up to Black Friday, the busiest shopping day of the year. But the sheer amount of compute capacity required in reaching out to millions of shoppers registers a hefty price tag for Drip. For Black Friday in 2019, the CRM firm took a novel approach: looking to optimize costs without compromising its customers’ connections to consumers, the company turned to Amazon Web Services (AWS) and Amazon Elastic Compute Cloud (Amazon EC2) Spot Instances for a solution.

Drip offers a bridge for small to midsize e-commerce merchants to engage their customers more effectively at every point of the customer journey through email automation workflows, timed communication blasts, and social media campaigns. Drip also helps businesses tailor the individual shopping experience through its customizable CRM platform. Jason Selby, Drip vice president of engineering, explains: “Our customers can re-create that high-touch, high-curated experience they have in store through our platform, which offers deep customization, including purchase history, browsing preferences, and recommendations. This touches the customer at the right points in their journey online in the same way that a sales associate in store would do.” And during retail’s busiest time of the year, Drip needs enough compute capacity to provide that connection for all the ways Drip’s customers seek to engage their own customers.

“On a typical week leading up to Black Friday, we send probably 10–15 million emails a day,” says Drip principal architect Dustin Blomquist. That’s up to four times Drip’s normal demand. Workload jumps like these require a huge amount of not just computing power but also flexible scalability from the instance fleet. Before 2019, Drip fulfilled its computing and scalability needs with a blend of Amazon EC2 Reserved Instances and Amazon EC2 On-Demand Instances to meet these peaks in workload. Reserved Instances enable businesses like Drip to reserve compute capacity ahead of time, whereas On-Demand Instances are available as needed to handle short-term, irregular workloads that cannot be interrupted—such as Drip’s increased workloads surrounding Black Friday. This solution worked but required significant upfront investment, and the expense limited Drip’s load testing and provisioning capacity.

To solve this problem, AWS proposed that Drip adopt Amazon EC2 Spot Instances, which let users take advantage of unused Amazon EC2 capacity in the AWS Cloud at up to a 90 percent discount from On-Demand Instance pricing, explaining that it held the potential to save costs and increase capacity with a relatively easy restructuring process. “The cost of adoption looked quite simple,” says Selby, “and the cost savings opportunity was great. So at that point we decided to invest the engineering time to change how our infrastructure was provisioned.”

The Drip team developed an innovative Amazon EC2 Spot Instances–backed environment wherein the company could capitalize on the savings of Amazon EC2 Spot Instances without compromising service. “As we got more comfortable with Amazon EC2 Spot Instances, we reconfigured our Reserved Instances,” says Selby. “Then we’d use Amazon EC2 Spot Instances, only adding On-Demand Instances when we needed to.” Drip followed Amazon EC2 Spot Instances best practices about diversifying its instance fleet to choose as many capacity pools with different instance types, sizes, and Availability Zones as could be supported by its workloads. Drip also launched Amazon EC2 Spot Instances with a capacity-optimized allocation strategy that would select the deepest Amazon EC2 Spot Instances capacity pools, significantly reducing interruptions. This process played a key role in increasing Drip’s confidence about maintaining Amazon EC2 Spot Instances capacity at scale during peak seasonal demands. Within weeks of migrating, Drip had more than 80 percent of asynchronous workloads switched over to Amazon EC2 Spot Instances.

In anticipation of its first Black Friday shopping crunch postmigration, Drip ran exhaustive load tests, approaching close to 100 million emails in a single day. And thanks to the affordable Amazon EC2 Spot Instances, Drip could essentially do more for less. It could perform these tests and add more compute as needed without worrying about costs increasing. “We overprovisioned our system so much that we were able to multiply our normal rate of email throughput leading up to Black Friday by a factor of six to seven,” says Blomquist. “We had never pushed our system anywhere near that hard.”

Having put the new Amazon EC2 Spot Instances–backed infrastructure through its paces, Drip was ready for the big day—and when it came, everything went smoothly. “Amazon EC2 Spot Instances were a massive win for us for Black Friday,” says Selby. “We saw the cheapest Thanksgiving we’ve had in several years from a raw dollars perspective.” All told, Drip saved more than $100,000—at least 65 percent in compute costs from the previous year—even while running more hours than before. “We ended up converting a few additional workloads post-Thanksgiving,” says Selby, “so we’ve probably pulled the costs down even further.”

The enormous cost savings had an indirect benefit on performance as well. “Because Amazon EC2 Spot Instances were less expensive,” says Selby, “we were much more comfortable leaving our system overprovisioned—with extra capacity not being used—so that when customer demand would arrive, we already had instances running and available to service that demand immediately rather than having to wait the 90 seconds or so it takes to spin up new instances as you scale.” Not needing to spin up instances translated into almost half the wait time for businesses to see their campaigns go from launched to in front of their customers. “For example,” explains Selby, “before, customers had to wait, say, 9 minutes for their 600,000 people to get their emails; now, we could get it done in 4.5 minutes.”

Drip is looking forward to discovering new ways it can continue to optimize with AWS. Drip uses Amazon Elastic Container Service (Amazon ECS)—a fully managed container orchestration service known for its security, reliability, and scalability—for some of its microservices and is intrigued by the prospect of uncovering more potential. Meanwhile, though Drip is already 85 percent switched over to Amazon EC2 Spot Instances, it is pursuing migrating the few remaining holdout workloads and autoscaling groups it has to Amazon EC2 Spot Instances as well. “A lot of our focus as we’re thinking about next Black Friday is around scaling other things and shifting other workloads to scale in smaller increments and scale more dynamically,” says Selby.

Drip used Amazon EC2 Spot Instances to save considerably on costs during the busiest time of the year. Now Drip has the opportunity to turn those savings into new solutions, new features, and stronger connections, helping businesses lead the way to more personalized and engaging customer journeys.